CVE-2026-41089 Netlogon RCE Hits Domain Controllers
Microsoft's Netlogon Remote Code Execution vulnerability (CVE-2026-41089, CVSS 9.8) is under active exploitation by threat actors targeting corporate networks worldwide. This critical flaw allows attackers to seize complete control of Windows domain controllers and gain SYSTEM privileges without prior authorization or user interaction. The Centre for Cybersecurity Belgium recently confirmed these in-the-wild attacks, which demonstrates an immediate and severe risk to enterprise infrastructure.
The widespread use of Netlogon in enterprise environments makes this vulnerability dangerous. Successful exploitation grants adversaries an uncontested foothold, enabling broad reconnaissance, privilege escalation, lateral movement, and identity compromise across an organization's most sensitive systems. This could also disrupt critical business operations.
Microsoft released patches for CVE-2026-41089 as part of its May 2026 Patch Tuesday. This update addressed 118 vulnerabilities, 16 rated as critical. The urgent deployment of these specific fixes is important for organizations to protect their core authentication systems from ongoing exploitation.
How is the Netlogon RCE vulnerability being exploited in the wild?
Threat actors exploit CVE-2026-41089 by sending specially crafted network requests to unpatched Windows domain controllers. This causes the Netlogon service to improperly handle incoming data, which enables arbitrary code execution on the target system. Its CVSS score of 9.8 shows the vulnerability's critical impact.
The exploit grants attackers full SYSTEM privileges on the affected domain controller, giving them complete administrative control over the compromised system. This malicious action requires zero user interaction and no prior authorization, allowing for silent and rapid compromise. The Centre for Cybersecurity Belgium issued an advisory confirming active exploitation, stressing that immediate defensive measures are needed to safeguard corporate infrastructure.
Microsoft provided fixes for Windows Server versions from 2012 onwards in its May 2026 Patch Tuesday release. Organizations that have not yet applied these updates remain vulnerable to these active attacks. Prompt application of these security patches is the primary defense against this critical vulnerability.
Remediation for CVE-2026-41089
System administrators must immediately deploy fixes for all affected Windows Server versions. This important step is the most effective way to prevent ongoing exploitation. Organizations should also consider isolating exposed domain controllers from untrusted networks to limit their attack surface while patches are applied.
Further protective measures include continuous monitoring of network traffic for unusual authentication requests. Security operations centers should regularly review Netlogon logs for anomalies that could indicate attempted or successful exploitation. Maintaining strict patch hygiene and active monitoring remain fundamental defenses against advanced threat groups using vulnerabilities like CVE-2026-41089.
Past incidents involving actively exploited Microsoft Exchange zero-days demonstrate the persistent threat posed by critical vulnerabilities in core enterprise services.
Dutch Authorities Dismantle Large-Scale Asocks Botnet Affecting 17 Million Devices
Dutch authorities, in a joint operation involving the Politie and the National Cyber Security Center (NCSC), have successfully dismantled a massive botnet known as Asocks. This botnet enslaved at least 17 million infected devices globally, using them for various malicious activities. The operation involved seizing more than 200 servers located in the Netherlands that served as the botnet's backend infrastructure.
The infected devices encompassed a broad spectrum, including computers, tablets, smartphones, and various IoT devices. The NCSC identified Asocks as a residential proxy service, which, while having legitimate uses, is frequently abused by threat actors. Reports from local news outlet NL Times corroborated the botnet's identity as Asocks.
Previous intelligence reports, such as HUMAN's Satori Threat Intelligence team's findings in April 2024, linked Asocks to a campaign dubbed PROXYLIB. This campaign involved infecting Android devices with proxyware from both LumiApps and Asocks, showing a history of malicious activity associated with this service. The law enforcement action involved seizing a subset of these servers from a hosting provider, which took the botnet offline.
Botnet Infection and Mitigation
Devices typically become part of a botnet when threat actors gain unauthorized access and install malware for remote control. This integrates the compromised device into a network used for cybercriminal activities. The scale of the Asocks botnet shows the pervasive nature of such infections.
To counter botnet malware, organizations and individuals should implement several security practices. These include keeping operating systems and software applications up-to-date with the latest security patches. Maintaining visibility of edge devices, such as routers, is also important for identifying and mitigating potential compromises.
Additional recommendations include using strong, unique passwords for all accounts and enabling two-factor authentication (2FA) wherever possible. Installing applications only from trusted sources and changing default passwords on new devices, particularly IoT devices, are important steps. Securing Wi-Fi networks with strong encryption protocols like WPA2 or WPA3 can further prevent unauthorized access and potential botnet recruitment.
Malicious Codex UI npm Package Steals OpenAI Refresh Tokens from 27,000 Developers
A malicious npm package, codexui-android, a popular remote web user interface for OpenAI Codex, has been discovered exfiltrating OpenAI refresh tokens from users. With an estimated 27,000 weekly downloads, this supply chain attack exposed a significant number of developers to persistent account takeover risks. Aikido Security researcher Charlie Eriksen made the discovery on May 27, 2026.
The attackers employed a deceptive strategy, developing a useful tool likely to establish a legitimate user base before initiating malicious activity. The important element of this attack is that the malicious code was not present in the public GitHub repository. Instead, it was found exclusively within the published npm package, allowing it to bypass standard source code audits.
The attack initiates immediately upon module load, with the dist-cli/index.js file importing a hidden script named chunk-PUR7OUAG.js. This script then checks for local credentials. If found, it launches a data exfiltration routine to steal access_token, id_token, account ID, and the important refresh_token from the auth.json file. The refresh token is dangerous as it typically does not expire, granting attackers indefinite access and impersonation capabilities.
Covert Data Exfiltration and Associated Android Apps
To evade detection, the exfiltrated data was sent to a server endpoint named sentry.anyclawstore. This endpoint was chosen intentionally to blend in with normal Sentry error-reporting telemetry. The hidden source map even contained a comment from the author: "Send tokens to our startlog endpoint (always)," indicating deliberate intent.
Further investigation revealed that the threat actor behind this package, operating under the developer identity BrutalStrike, also targeted Android mobile devices. BrutalStrike has published applications on the Google Play Store, including a paid productivity app called codex.app and another named "OpenClaw Codex Claude AI Agent." Both these applications were found to contain the same malicious infrastructure.
These Android apps initially bypassed Google's pre-publish security scans due to their clean initial APK file. Once installed, the app extracts a Termux-derived Linux userland into private storage and launches Node.js using PRoot. It then executes a command to install the latest version of the npm package: pnpm add codexui-android@latest. The exfiltration functionality has been active since version 1.0.0 of the package.
Critical Langroid Prompt Injection Vulnerability Exposes LLM Applications to RCE
Researchers from Carnegie Mellon University (CMU) and the University of Wisconsin-Madison (UW-Madison) have identified a critical Remote Code Execution (RCE) flaw within the Langroid Python framework, specifically affecting Large Language Model (LLM) applications. This vulnerability, which achieved a maximum CVSS score of 9.8, presents a significant risk to database servers by allowing prompt injection attacks. Developers are urged to upgrade their installations immediately.
The core of the problem resides within the framework's SQLChatAgent component. While designed to execute database queries generated by an underlying language model, the component can be manipulated by malicious users through prompt injection. If the database role associated with the agent possesses elevated administrative privileges, the consequences can be severe.
An attacker can force the system to execute dangerous dialect-specific primitives. For example, on a PostgreSQL backend, an attacker could trigger commands like COPY FROM PROGRAM. This action facilitates full RCE on the underlying database host, granting the attacker extensive control. This type of vulnerability shows the emerging security challenges in the rapidly evolving field of LLM-powered applications.
Security Impact and Patch Availability
Successful exploitation of this flaw carries high security implications. Adversaries could execute arbitrary system commands using the database's local privileges, potentially leading to a broader compromise of the network. Furthermore, attackers could silently exfiltrate sensitive corporate data from the server or maliciously modify and delete critical database tables. The ability to pivot through the network from a compromised database server makes this an important entry point for sophisticated attacks.
Fortunately, the Langroid development team has addressed this dangerous RCE bug. A security patch is available in version 0.63.0 and all newer releases of the framework. This update introduces a strict SELECT-only allowlist parsed by sqlglot and implements a dialect-aware blocklist to prevent dangerous operation patterns. Users can manually restore the old behavior via a configuration flag in trusted environments, though this is not recommended for most deployments.
Addressing vulnerabilities in development frameworks is as critical as patching operating systems. Lessons from past critical Microsoft Defender zero-days show the importance of rapid patching across the entire software supply chain.
Hard-Coded Password Exposes Eppendorf BioFlo 320 Bioreactor Systems
An urgent industrial control security warning has been issued for laboratory facilities concerning a critical flaw in the Eppendorf BioFlo 320 bioreactor platform. This high-severity vulnerability (CVE-2026-7251, CVSS 9.8) exposes these systems to unauthorized manipulation of sensitive biochemical processes. Lab managers must inspect their device configurations to prevent potential safety incidents.
The software defect originates from a poorly secured remote management tool. The underlying system relies on an exposed Virtual Network Computing (VNC) architecture that uses a hard-coded password. Official documentation confirms: "The affected product is vulnerable due to VNC server using a hard-coded password." This access mechanism lacks encrypted network interactions, further reducing its security posture.
If an attacker identifies the network address of a target system, they can exploit this default credential to gain unauthenticated administrative authority. The vulnerability report explicitly states: "Once connected, the attacker would have full access to all control panel features for the BioFlo 320." This level of access could enable important changes to experiments, potentially compromising research integrity or causing hazardous conditions in laboratory settings.
Remediation for CVE-2026-7251
Eppendorf, the manufacturer, has developed an update to eliminate this threat. The newly released Version 5.0 software patch safely disables the vulnerable remote control protocol. Importantly, all systems originally shipped with this feature deactivated by default. Users could only activate the module manually at the physical workstation tower.
Applying the permanent fix resolves the Eppendorf bioreactor security flaw, protecting important laboratory equipment. Administrators should download and apply the latest Version 5.0 software package without delay. Furthermore, security teams should verify that local user role protections adequately restrict configuration changes to trusted supervisors, adding another layer of defense against unauthorized access.
Technical Takeaways
- Active Exploitation of Core Enterprise Systems: Microsoft's Netlogon RCE, CVE-2026-41089 (CVSS 9.8), is actively exploited to gain SYSTEM privileges on Windows domain controllers, requiring immediate patching of Windows Server versions from 2012 onwards.
- Large-Scale Botnet Disruption: Dutch authorities dismantled the Asocks botnet, which comprised 17 million infected devices, including Android and IoT devices, linked to the PROXYLIB campaign.
- Supply Chain Attacks Targeting AI Development: A malicious
codexui-androidnpm package, with 27,000 weekly downloads, was found stealing persistent OpenAI refresh tokens via a covert supply chain attack, also impacting related Android applications. - Prompt Injection Risks in LLM Frameworks: A critical Langroid Python framework vulnerability (CVSS 9.8) allows RCE via prompt injection in the
SQLChatAgentcomponent, enabling arbitrary command execution on database hosts without authentication. - ICS/OT Hard-Coded Credential Vulnerability: The Eppendorf BioFlo 320 bioreactor platform has a critical flaw, CVE-2026-7251 (CVSS 9.8), due to a hard-coded VNC password, allowing unauthenticated administrative access to industrial control functions.
